Thanks. I figured the Nyquist rate would come into play here. What I figured I'd do is first get the system running with the best sampling rate I could reasonably afford, and then work out the error statistics later on in my analysis.
On Wed, Oct 19, 2011 at 8:42 AM, Manuel Wolfshant <wo...@nobugconsulting.ro> wrote: > On 10/19/2011 03:40 PM, Christian Convey wrote: >> >> Thanks for the ideas. >> >> You're right - I *am* looking for total energy expenditure (Joules, >> Watt-hours, etc.) I have no specific desire to do that integration in >> my own code. >> >> As far as sampling frequency, here's the deal: I'm working on a >> system to estimate the power draw associated with a run (or set of >> runs) of an arbitrary program on my Linux computer. Some of the >> program runs will be brief, but I'm not sure how brief we're talking. >> I'm no statistics expert, but I believe my power-draw estimates will >> get pretty inaccurate unless the sampling rate is markedly higher than >> the program's running time. I *can* limit my studies to programs that >> run for a long time (so as to reduce the relative error stemming from >> a slow-sampling power meter), but that would put an unwelcome >> limitation on my research. > > http://en.wikipedia.org/wiki/Sampling_rate > mind the Nyquist–Shannon sampling theorem > > > _______________________________________________ > Nut-upsuser mailing list > Nut-upsuser@lists.alioth.debian.org > http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/nut-upsuser > _______________________________________________ Nut-upsuser mailing list Nut-upsuser@lists.alioth.debian.org http://lists.alioth.debian.org/cgi-bin/mailman/listinfo/nut-upsuser